[v2] md: fix that MD_RECOVERY_RUNNING can be cleared while sync_thread is still running

[PATCH -next v2 0/6] md: fix that MD_RECOVERY_RUNNING can be cleared while sync_thread is still running

Posted by Yu Kuai 2 years, 8 months ago

From: Yu Kuai <yukuai3@huawei.com>

Changes in v2:
 - rebase for the latest md-next

Patch 1 revert the commit because it will cause MD_RECOVERY_RUNNING to be
cleared while sync_thread is still running. The deadlock this patch tries
to fix will be fixed by patch 2-5.

Patch 6 enhance checking to prevent MD_RECOVERY_RUNNING to be cleared
while sync_thread is still running.

Yu Kuai (6):
  Revert "md: unlock mddev before reap sync_thread in action_store"
  md: refactor action_store() for 'idle' and 'frozen'
  md: add a mutex to synchronize idle and frozen in action_store()
  md: refactor idle/frozen_sync_thread() to fix deadlock
  md: wake up 'resync_wait' at last in md_reap_sync_thread()
  md: enhance checking in md_check_recovery()

 drivers/md/dm-raid.c |   1 -
 drivers/md/md.c      | 124 +++++++++++++++++++++++++++++--------------
 drivers/md/md.h      |   5 ++
 3 files changed, 88 insertions(+), 42 deletions(-)

-- 
2.39.2

Re: [PATCH -next v2 0/6] md: fix that MD_RECOVERY_RUNNING can be cleared while sync_thread is still running

Posted by Yu Kuai 2 years, 8 months ago

Hi,

在 2023/05/29 21:20, Yu Kuai 写道:
> From: Yu Kuai <yukuai3@huawei.com>
> 
> Changes in v2:
>   - rebase for the latest md-next
> 
> Patch 1 revert the commit because it will cause MD_RECOVERY_RUNNING to be
> cleared while sync_thread is still running. The deadlock this patch tries
> to fix will be fixed by patch 2-5.
> 
> Patch 6 enhance checking to prevent MD_RECOVERY_RUNNING to be cleared
> while sync_thread is still running.

Any suggestions on this patchset? I already sent regression test
for the deadlock problem for both raid10 and raid456.

Thanks,
Kuai
> 
> Yu Kuai (6):
>    Revert "md: unlock mddev before reap sync_thread in action_store"
>    md: refactor action_store() for 'idle' and 'frozen'
>    md: add a mutex to synchronize idle and frozen in action_store()
>    md: refactor idle/frozen_sync_thread() to fix deadlock
>    md: wake up 'resync_wait' at last in md_reap_sync_thread()
>    md: enhance checking in md_check_recovery()
> 
>   drivers/md/dm-raid.c |   1 -
>   drivers/md/md.c      | 124 +++++++++++++++++++++++++++++--------------
>   drivers/md/md.h      |   5 ++
>   3 files changed, 88 insertions(+), 42 deletions(-)
>

Re: [PATCH -next v2 0/6] md: fix that MD_RECOVERY_RUNNING can be cleared while sync_thread is still running

Posted by Song Liu 2 years, 8 months ago

On Wed, Jun 7, 2023 at 7:41 PM Yu Kuai <yukuai1@huaweicloud.com> wrote:
>
> Hi,
>
> 在 2023/05/29 21:20, Yu Kuai 写道:
> > From: Yu Kuai <yukuai3@huawei.com>
> >
> > Changes in v2:
> >   - rebase for the latest md-next
> >
> > Patch 1 revert the commit because it will cause MD_RECOVERY_RUNNING to be
> > cleared while sync_thread is still running. The deadlock this patch tries
> > to fix will be fixed by patch 2-5.
> >
> > Patch 6 enhance checking to prevent MD_RECOVERY_RUNNING to be cleared
> > while sync_thread is still running.
>
> Any suggestions on this patchset? I already sent regression test
> for the deadlock problem for both raid10 and raid456.

Sorry for the delay. I will look into this soon.

Thanks,
Song